Evaluating named entity recognition and disambiguation in news and tweets